Skip to main content

Mete Ahishali: Revolutionary sparse representation techniques propel the efficiency and accuracy of artificial intelligence

Tampere University
LocationKorkeakoulunkatu 1, Tampere
Hervanta campus, Tietotalo, auditorium TB109 and remote connection
Date12.8.2024 9.00–13.00
Entrance feeFree of charge
A person wearing a cap. Forest and lake in the background.
Photo: Fahad Sohrab
Recent breakthroughs in the field of Machine Learning have accelerated research exploring sparse representation in pattern recognition. Sparse representation, also known as sparse coding, refers to the compact yet sufficient depiction of data where only the most descriptive components are preserved. In his doctoral dissertation, Mete Ahishali proposes novel machine learning methods that harness sparse representations achieving state-of-the-art performances across a wide range of pattern recognition applications.

There has been a growing interest towards data-driven Machine Learning in the era of Artificial Intelligence. In his doctoral dissertation, Mete Ahishali focuses on developing advanced machine learning methodologies leveraging sparse representations.

“These methods have shown significant improvements and better computational efficiency than conventional machine learning methods. Sparse representation of data can be defined by exploiting the most important information, that is similar to predicting the key pieces for a given complex puzzle”, Mete Ahishali says.

Especially in challenging decision-making environments, such as when the data is limited or noisy, this approach enables the artificial intelligence to perform faster and more efficiently without compromising accuracy. Mete Ahishali shows that the cutting-edge technologies proposed in his thesis have the potential to enhance various tasks performed by artificial intelligence. For example, these technologies have been evaluated in different applications from computer-aided medical diagnosis to facial recognition, image acquisition/reconstruction, analysing satellite imagery in environmental monitoring, and even autonomous vehicles. Additionally, proposed machine learning methods have provided better quality and improved speed in Magnetic Resonance Imaging (MRI).

“A new type of artificial neural networks is developed in this thesis to learn and adapt over time. This process is inspired by the biological neural model and provides a better approximation compared to the existing ones allowing more accurate and efficient AI technologies in future applications”, Mete Ahishali states.

Overall, these groundbreaking contributions open new opportunities for innovation and discoveries of “efficient AI” leveraging sparse representations in order to solve real-world problems.

Mete Ahishali has conducted his doctoral research at Signal Analysis and Machine Intelligence (SAMI) research group at Tampere University.

Public defence on Monday 12 August

The doctoral dissertation of M.Sc. Mete Ahishali in the field of Computing and Electrical Engineering titled Advanced Machine Learning for Sparse Representations in Pattern Recognition Applications will be publicly examined at the Faculty of Information Technology and Communication Sciences at Tampere University at 12:00 on Monday 12 August 2024. The venue is auditorium TB109 of the Tietotalo Building (Korkeakoulunkatu 1, Tampere). The Opponent will be Professor Ghassan AlRegib from Georgia Institute of Technology. The Custos will be Professor Moncef Gabbouj from the Faculty of Information Technology and Communication Sciences at Tampere University.

The doctoral dissertation is available online

The public defence can be followed via remote connection